Design And Implementation Of Lateral Control System For Automatic Driving In Closed Park

As the core of the autonomous driving system,vehicle control is responsible for maneuvering the vehicle to travel along the reference path accurately and smoothly.The closed park environment is most suitable for autonomous driving technology because of its simple environment,high degree of structure and frequent driving routes.In the closed park,the speed is always limited under 30-40km/h,and the power requirement of the vehicle is low,so there are always low-speed battery cars applied in this environment,which limits the power and calculate ability of the equipment loaded in the car and could not hold very complicated algorithms.Meanwhile,the curvature radius of the roads in the closed park can be 5m at least where the control accuracy of the traditional algorithms drops significantly.Among the lateral path following strategies,preview following model is widely applied because of its simple structure and making full use of future road information.The fuzzy control algorithm is constructed based on fuzzy rules,especially suitable for the control of nonlinear systems,which is easy to express human's existing control experience,and does not need to establish an accurate mathematical model of the controlled object,and has strong robustness.Therefore,in this thesis the lateral preview fuzzy control algorithm is studied,and the following aspects are mainly included:(1)An autonomous driving vehicle control system integrating obstacle detection system,navigation system,control system and execution system was designed with the lateral control as the main research purpose.And based on an electric patrol car,the hardware structure of the control system was established with the industrial computer as the core,and the electronic transformation plan of the executive organization was formulated.The navigation scheme with two-antenna RTK-GPS was determined by comparison.The assembly scheme of LiDAR was established according to the distribution characteristics of VLP-16 LiDAR and the detection range of obstacles.(2)The lateral preview fuzzy control algorithm based on independent preview distances has poor tracking of the reference path with small radius of curvature under the condition of low speed.Aiming for this problem,the optimal preview distances of the lateral error with different speed conditions and curvature radius were obtained through the simulation based on the lateral kinematic vehicle model and the calculation method of curvature radius.Then,the fuzzy neural network based on the Takagi-Sugeno fuzzy logic was applied to learn the simulation data where the speed and the curvature radius of the present position is set as the input.The network is then used for the adaptive generation of lateral distance error preview distance..(3)A co-simulation model based on CarSim and Simulink was built.Then a continuous curvature with small changing radius was designed where in the radius range of 5 to 15 m.The simulation for both the original lateral control algorithm and the improved one with adaptive preview distance generation method was tested.The variance of the present lateral error generated by the improved method reduces to 0.12 times the unimproved ones.The result shows that the improved algorithm has better tracking effect for the reference paths with the radius of curvature smaller than 9m.(4)An autonomous driving software system was developed.First,the architecture of system was designed.Secondly,the appropriate navigation data protocol and the calculating methods of the distance and the bearing angle were selected.Then the communication protocols of the control system and the obstacle avoidance system were formulated,and the corresponding execution logic and the software flow were designed accruing to the system operation targets.Finally,an intuitive human-computer interface was designed.